基于H.264的运动矢量集自适应快速搜索算法

A fast searching algorithm using adaptive motion vector aggregate based on H.264

为了减少H.264编码器中运动估计模块的计算复杂度,提出了一种基于H.264的运动矢量集自适应的EPZS改进算法。该算法充分利用运动矢量的时空相关性和中心偏置特性,通过对运动类型判定,自适应选取初始预测矢量集和搜索模式,并根据运动类型和匹配误差的空间方向特性,提出了一种新的自适应阀值和一种新的非对称十字-六边形混合搜索模板。实验结果表明,该算法与EPZS算法相比,在编码性能相当的情况下,运动估计时间平均节省了18.18%。

In order to reduce the computational complexity of motion estimation in encoder of H. 264, an improved EPZS algorithm with the adaptive motion vector aggregate based on H. 264 is presented in this paper. This algorithm makes full use of the spatial-temporal relativity and the central-biased characteristic of the motion vector , adopts the adaptive selecting initial prediction of motion vector aggregate and the searching modes by judging motion types, according to the motion types and matching error. A new adaptive adjusting threshold and a novel unsymmetrical-cross hexagon mixing searching pattern are proposed. Experimental results show that, compared to EPZS algorithm, the algorithm can save about 18.18% of motion estimation time,but almost no change in the encoding behaviour.